40 Corning) and 10% (w/w) curing agent (Dow Corning) was poured over the SU-8
41 molds, degassed and incubated at 65 degree overnight. Polymerized PDMS was
42 peeled off from the mold activated by incubation for 1 min in an oxygen plasma oven
43 (Diemer Femto) and bound to a 50 × 75 × 0.4 mm ITO glass (Delta Technologies).
44 Inlets and outlets were punched using 0.5 mm diameter biopsy punches (Harris Uni-
45 Core) for electrodes and 0.75 mm diameter biopsy punches for the rest. The
46 channels were first flushed by Aquapel (PPG Industries) and, subsequently, by
47 HFE7500 oil (3M).
48
49 Cell cultivation and encapsulation
50 Her2 Hybidoma cells (ATCC® CRL-10463) were grown in complete DMEM medium
51 (Gibco), Jurkat cells (ATCC® TIB-152) were grown in RPMI medium (Gibco), both
52 supplemented with 10% FBS. Hybridoma cells were harvested, stained by Calcein-
53 AM (Lifetechnologies) and Calcein Violet (E-bioscience), respectively, at room
54 temperature for 45 min, washed by PBS twice to remove free dye in the media , and
55 re-suspended in free style media (Gibco) supplemented with 1 mg/ml xanthan gum
56 (Sigma) to prevent cell sedimentation during encapsulation. Subsequently, green and
57 violet cells were mixed equally at a final concentration of 1.5 × 106 cells/ml and
58 injected at a flow rate of 1000 µl/h into the droplet generation chip. Droplets were
59 generated by flow focusing this continuous phase using Novec HFE7500 oil,
60 containing 5% PEG surfactant3 (custom synthesized at Sigma Aldrich), at a flow rate
3
61 of 4000 ul/h. Emulsions were collected in a collection tube (cryotobube, Nunc) which
62 was treated with Aquapel (PPG industries) and, subsequently, rinsed by HFE7500
63 oil.
64 Sorting and Imaging
65 Emulsions were re-injected using an electro-osmotic pump (Nano Fusion
66 Technologies) at a flow rate of about 60 µl/h. Oil with 0.5% and 0.25% of PEG
67 surfactant were loaded in syringes individually and injected by Harvard Apparatus
68 PHD 2000 syringe pumps at a flow rate of 400 µl/h (Fig.2B, (a)&(c)) and 600µl/h
69 (Fig.2B, (d)) respectively . A refilling pump was connected with outlet E (Fig.2B, (e))
70 to withdraw all of the droplets that did not trigger sorting to the waste syringe at a flow
71 rate of 760 µl/h. Droplet sorting videos were acquired at ~500 frames per second. A
72 customized LabVIEW sorting program was used to control the droplets sorting. The
73 positive droplets were collected in the collection chip (Fig. 2F-H) and the trapping
74 events were monitored on a cell imaging device (CytoMate Inc.). The collection was
75 finished when all of the traps were occupied. Subsequently, the collection chip was
76 rinsed with oil containing 0.25% PEG surfactant to remove un-trapped droplets.
77 Sorting enrichment was determined by automated scanning of the entire collection
78 chip at 10-fold magnification using an inverted fluorescence microscope (Nikon
79 eclipse Ti), equipped with a motorized stage and a Hamamatsu Digital camera.
80 Images were stitched, processed and analyzed using ilastik (ilastik.org) and ImageJ.
81
4
82 Table S1: Sorting results. ND = not detectable
Dual color droplets
Single
color
droplets
Empty
dropletsCell type
2 cells 3 cells ≥4 cells ≥1 cell 0 cells
hybridoma
cellsNumber 379 83 4 3138 ND
Before
sorting
(PMT data) Percentage 10.5% 2.3% 0.1% 87.1% ND
Number 5 2 1 57 161Before
sorting
(collection
chip)
Percentage 2.2% 0.9% 0.4% 25.2% 71.2%
Number 191 47 5 6 0After
sorting
(collection
chip)
Percentage 76.7% 18.9% 2% 2.4% 0
Jurkat cells Number 2291 535 69 16868 NDBefore
sorting
(PMT data)Percentage 11.6% 2.7% 0.4% 85.4% ND
Number 14 10 2 177 291Before
sorting
(collection
chip)
Percentage 2.8% 2.0% 0.4% 35.8% 58.9%
Number 402 32 0 26 4After
sorting
(collection
chip)
Percentage 86.6% 6.9% 0.0% 5.6% 0.9%
83
5
8485 Figure S1. Flowchart summarizing the logic of the LabVIEW control software 86 programmed at EMBL, Heidelberg. This algorithm runs in parallel for both of the PMT 87 channels (one for each colour) and detects peaks in the signal values. This allows cells 88 within droplets to be detected and for a sorting decision to be made for each passing 89 droplet based on the intensity of the signal, the number of peaks detected, the width of 90 the overall peak and the spacing between droplets that contain at least one cell. This 91 software and a user manual can be freely downloaded for academic use at 92 www.merten.embl.de/index.html.
93
6
94
95 Fig. S2. Schematic of the optical setup. The fluorescence-based sorting setup 96 uses diode lasers with excitation wavelengths of 405 nm (Calcein Violet), 488 nm 97 (Calcein-AM) and 561 nm (optional third laser for assay readouts). Emission signals 98 are detected using PMTs with a 450 nm band-pass filter (blue), a 521 nm band-pass 99 filter (green), and a 610 nm longpass filter (red). Sorting signals are processed using
100 LabVIEW software running on a FPGA card triggering a high voltage amplifier. 101 Imaging is performed using an inverted microscope equipped with a high speed 102 camera. 103
104105
7
106 Fig. S3. Example of the signal peaks in one droplet. The zoom in (inset) reveals a 107 jigsaw shape of the signal at low intensity, thus making the use of inflection points for 108 the detection of peaks impossible. 109
110111112 Fig. S4. Signal variation of Calcein-AM and Calcein-violet stained cells inside 113 droplets. (A) Droplet showing one green peak and two overlapping violet peaks, 114 corresponding to a clump of 2 violet cells, with a valley between the two peaks above 115 a value of 0.5 fluorescence units. This value is higher than the green peak of another 116 droplet (B) hosting exactly one green and one violet cell. Therefore using static 117 thresholds (solid black lines) is not sufficient to accurately detect the number of 118 encapsulated cells. However, when specifically detecting drops in the fluorescence 119 signal exceeding the maximum noise (red dots), the number of peaks can be 120 correctly determined, independently of the peak intensities.121
8
122
123
124 Figure S5. Leakage of Calcein Violet from cells encapsulated into droplets. (A) 125 Zoom in of the fluorescence signals over 8 hours incubation at room temperature. The 126 strongly decreased scale of the Y-axis (from 0 to 0.05 A.U.) allows illustrating the 127 increase in the droplet signal (wide peaks), but requires cropping of the cell signals 128 (narrow subpeaks with intensities as shown in (B)). (B) Time course of fluorescence 129 signals of droplets hosting Calcein Violet-stained hybridoma cells. After incubation for the 130 indicated time periods off-chip, the droplets were reinjected into the sorting device and 131 the fluorescence signals were determined in the detection channel using a PMT.. (C) 132 Fitted LOESS smoothing line of droplet fluorescence intensities (turquoise line), 133 individual data points (turquoise circles) and confidence (grey shades) of the droplet 134 signals. (D) Fitted LOESS smoothing line of cell fluorescence intensities (red line), 135 individual data points (red circles) and confidence bands (grey shades) of the cell signals. 136 (E) Intensities of cell and droplet signals plotted at the same scale.137
9
138139140 Figure S6. Leakage of Calcein Violet from cells cultivated in glass bottom wells. (A) 141 Calcein Violet-stained hybridoma cells were incubated for the indicated time periods (X-142 axis) and Fluorescence intensities of the media supernatants over time was determined 143 by imaging (manually selected samples analyzed for their intensity using ImageJ). (B) 144 Fluorescence intensities of the calcein violet-stained hybridoma cells over time. (C) 145 Intensities of the cell and droplet signals plotted at the same scale. Solid lines = fitted 146 LOESS smoothing line of fluorescence intensities; open circles = individual data points; 147 grey shades = confidence bands.148
10
149
150 Fig. S7. Efficiency of the sorting process for droplets hosting differently 151 stained Her2 Hybridoma cells. Blue fluorescence of droplets captured in the 152 collection chip before (A) and after (C) sorting. Green fluorescence of droplets 153 captured in the collection chip before (B) and after (D) sorting.154
155156157
11
158 Fig. S8. Fluorescence analysis of the droplets detected by PMT. (A) Two 159 dimensional dot plot of fluorescence signals of droplets. The red arrow indicates one 160 example of the dual color droplet with two cells. (B) Dot plot showing violet and green 161 signals of the droplets. (C) Droplet occupancy before sorting .
162
12
163
164165 Fig. S9. Efficiency of the sorting process for droplets hosting differently 166 stained Jurkat cells. Blue fluorescence of droplets captured in the collection chip
13
167 before (A) and after (E) sorting.Green fluorescence of droplets captured in the 168 collection chip before (B) and after (F) sorting. Bright field images of droplets 169 captured in the collection chip before (C) and after (G) sorting. Merged blue, green 170 and bright field images before (D) and after (H) sorting. (I) Droplet occupancies in 171 the collection chip before (top) and after (bottom) sorting.172173
174 Supplementary movie S1. Spacing of reinjected droplets upstream of the
175 sorting junction (visualized using a 2-fold objective).
176
177 Supplementary movie S2. Droplet sorting (visualized using a 40-fold objective). 178 The fluorescence signals of each droplet are measured in real time at 100 kHz and 179 processed. Droplets hosting exactly one green and one violet cell are actively pulled 180 into to the collection channel by switching on the electrodes. For all droplets with 181 undesired occupancies, the electrodes remain switched off and the droplets follow 182 the main flow into the waste channel. 183184 Supplementary movie S3. Leakage of Calcein Violet from cells cultivated in glass 185 bottom wells. Time lapse imaging of cells stained with Calcein Violet over a time period 186 of 9.5 hours with 30 minute intervals.
